Proposed grouping above material and immaterial entity
Opened this issue · 16 comments
It's not clear what the ME vs IE distinction buys us in COB. What would the implications of having a grouping "physical entity" that combines all current ME and IE children, with ME and IE going away (and continuing to live in BFO of course)
There would be no loss of information. the has_quality some mass parent would propagate down to the children of ME.
This would simplify many things. Currently scientific terminology is imprecise as to whether it is talking about an ME or IE. For example, in GO we have terms like nuclear lumen, these are treated as MEs. In AOs it is often arbitrary whether one chooses to model some things as spaces or the contents of those spaces. In geography/earth science there is no real ontological commitment on the part of scientists on whether things like areas, zones, etc denote IEs or MEs.
Of all the top-level classes, 'material entity' has the clearest differentiation conditions. Material entities have mass and a single location in space at a given time, that can only overlap with other material entities in limited ways.
Using material entities helps clarify the differentiation conditions for most of the other top-level classes: we can distinguish processes by their participants (locations and times), immaterial entities by their material boundaries, characteristics by their bearers, information entities by (sets of) their material concretizations (admittedly still difficult), etc.
Arguably, all these things are physical entities: space, time, energy, processes, information, characteristics.
If you need an encompassing class in needed, you could use 'spatially extended entity'. This could be a super class, or (perhaps) use an axiom to infer subclasses.
@wdduncan - I'm not so concerned with the name of whatever the union of the two concepts is. 'spatially extended entity' is perhaps fine though I see no a priori terminological reason why processes or even some characteristics may not be considered to be spatially extended entities
@bpeters42 and @jamesaoverton +1 to basic things in physics: but find me a physicist who deals with 'immaterial entities'.
We have discussed this a bit off-ticket, a brief summary is that my position is that IEs are frequently over-used and are often not necessary (see GO for example, which manages fine). However, insofar as IEs are used within a domain such as anatomy, it is pragmatic to group like with like, rather than carving in two, prioritizing a physics/philosophy split over a pragmatic user-centered one. However, the whole issue may go away if we can convince the main OBO AOs not to use IEs, and not make everything material as GO does.
processes or even some characteristics may not be considered to be spatially extended entities
By "spatially extended" I mean the typical 3 spatial dimensions present in classical/Newtonian physics. This would exclude processes. So, I'm not sure what you mean by processes being spatially extended.
The point about characteristics is debatable. But do we consider characteristics, such as color, to be spatially extended, or that the objects that have the characteristics are spatially ended? I think the latter makes more sense.
Colors also characterize material objects, but that doesn't necessarily imply that colors are material objects. Or does it?
Convincing AOs not to use IEs seems like a difficult task to me. Plus, I'm not sure how effectively you refer to things like eye sockets and ear canals.
@bpeters42: you are correct. Drop the not. I intended: "the whole issue may go away if we can convince the main OBO AOs not to use IEs, and make everything material as GO does"
Convincing AOs not to use IEs seems like a difficult task to me
Perhaps not, the majority of AOs that serve a major use community may prefer a solution that does not expose their users to things like 'immaterial anatomical entity'.
Plus, I'm not sure how effectively you refer to things like eye sockets and ear canals.
These are both material even in BFO-centric AOs. But if you mean concepts like the space bounded by the eye socket or the space surrounded by the ear canal, where truly required these classes could denote the respective material contents of said spaces. This is what GO does. Lysosomal lumen is a material entity that has material properties such as acidity.
@cmungall Having some axioms would help. Typically, we say that an eyeball is located in the eye socket (in the sense of 'eye socket' being a cavity) or that fluid is located in the ear canal (again understanding 'ear canal' in the sense of space in inside the ear). Would the eyeball now be part of the eye socket and the fluid be part of the ear canal?
Also, how well does this approach generalize. For example, see the discuss on geographic location (#91). Is Buffalo part of or located in NY State? Is my home located in or part of NY State?
If we take geographic locations to be sites, it seems to fine to assert that Mount Marcy (the tallest mountain in NY) is part of the portion of Earth that is located within the borders of NY State. But would this approach now dictate that Mount Marcy is part of NY State?
@bpeters42: makes sense. I would not object to a class labeled 'location' as a top level in COB.
@wdduncan - see uberon for axioms. ear canal = external acoustic meatus. creumen located-in EAM. outer epithelial layer of tympanic membrane partof EAM. No need to instantiate IE classes
The reference to UBERON confuses me a bit. You reference the external acoustic meatus, the physical structure of the tube. But, under this proposal, what is the distinction between the assertions x located in y and x part of y?
I've read the definitions for both located in and part of (as defined in RO). To understand, part of you have to understand a bit of mereology. But, intuitively, most users (ASIK) have been using located in for cases in which an entity is contained in some kind of space (such an IE or site); e.g. I am currently located in my living room, but I am not part of my living room.
UBERON, defines endolymph (the fluid in the inner ear) as located_in some 'endolymphatic space'. Which make good sense to me, since endolymphatic space is an anatomical cavity, and things are contained in cavities. But under this proposal, material entities would now be located in other material entities (I realize that we say things like this when speaking loosely). So, would an axiom such as hand located in arm be valid?
To reiterate my question about how well this approach would generalize:
Would location be a super class of geographic location (see #91 )? Geographic locations are currently IEs, but probably should be sites (IMHO). There is as sense in which a geographic location is material entity (e.g., the physical stuff that makes up area). But there is a non-material sense too (e.g., a geographic area is define by fiat things such borders). Under this proposal, how would we distinguish between these two senses if all geographic locations are material entities?
If location is a top level entity in COB, how would it be defined? If we define it in terms of positions relative to other entities, the issue of IEs and/or sites, there will still be the site that exists between the two entities being related.
I thought that my proposal for spatially extended entity was straightforward and intuitive. Both MEs and IEs are extended in 3D space. The intuitive distinction located in and part of remains intact, with no need to redefine things like anatomical cavities (e.g., endolymphatic space) or people being located in their homes in soley materialistic terms. I agree the label is somewhat ugly.
I just wanted to remind folks that these are discussions that went on for over 5 years earlier in the creation of CARO and in Uberon, as well as the many other anatomy ontologies. If there are decisions made here they will need to be vetted by the community of AO developers in a robust manner, who are probably not tracking COB. I also agree that we will need a way to ensure that the ontologies as disseminated don't confuse end users with unnecessary upper level components (not to say that the axioms can't be made/aligned, only that dissemination needs to be carefully considered). This is a robust and longstanding OBO community and they have specific needs that have been aligned pretty well over the years.
At the heart of it, anatomy is fundamentally about defining location relative to material entities and "spaces" that are defined by other physical entities. It is incorrect to call them material, though they may be filled with other material entities such as fluids. (will have to do a deeper dive on 'spatially extended entity')
As a reminder, this issue is about material vs. immaterial entities beyond anatomy. There were many objections in the discussion yesterday about environmental niches, geographical entities, and the like. So let's talk separately about what anatomy ontologies need, vs. what COB should provide in general.
I still think that a relation tying a 'location' to 'material content in location' (which we would want in COB in general) has a chance to address the needs of AO folks too.
On the proposed grouping term: I'm not sure what the need is for grouping at the top level. A proposal for grouping for anatomy (anatomical_entity) is here: #99 (comment). Is this not sufficient for COB?
I want to push back a bit against Chris' first assertion here:
Currently scientific terminology is imprecise as to whether it is talking about an ME or IE in GO we have terms like nuclear lumen, these are treated as MEs. In AOs it is often arbitrary whether one chooses to model some things as spaces or the contents of those spaces. In geography/earth science there is no real ontological commitment on the part of scientists on whether things like areas, zones, etc denote IEs or MEs.
The lumen of an (intracellular) organelle is filled with material that is part of the organelle and by extension, part of the cell. I don't know of any use case for referring to the space occupied by the contents or bounded by, for example, the organelle membrane. In this case, it makes sense in GO to use the lumen term to refer to the contents and this will correspond to the vast majority of usage of the term lumen in the literature. By contrast, the contents of the lumen of the gut or of a heart cavity are not part of the gut or heart respectively. In both cases we may want to refer to the shape of the lumen/cavity independently, e.g. to record a phenotype. In these case we are not referring to the shape of gut contents or some portion of blood.
The distinction I'm making here is, I think, very important for partonomy. In the GO case, we want the transitive chain of part relations to extend down from cell to the organelle lumen (understood as contents) and its parts. In the case of the contents of the gut or of a heart cavity we do not and we need a separate relation. I have long favoured located_in as a more general relation to part_of to record this - although we really need to use continuant_part_of for this to work. located_in can also be used to relate non-resident immune cells to a tissue, and foreign objects to cells or tissues.